In this study, we assessed the general physical problem-solving abilities of two closely re-lated Darwin’s finch species: the small tree finch and the woodpecker finch using the two-trap tube task with a pre-inserted tool that could be operated by both species. Furthermore, we assessed the effect of tool-using experience on general physical cognitive abilities in woodpecker finches and controlled for the effect of the unknown cognitive load imposed by the operation of a freely manipulable stick tool and a pre-inserted raking tool.
Figure 2.3 Performance of all birds in experiment 3 as percent trials correct per block.
Transfer performance for purpleblack is shown on the far right. The dotted line signifies chance performance (50%).
The comparison of woodpecker finches and small tree finches in Experiment 2 did not provide evidence that woodpecker finches, excel at solving the two-trap tube task and by ex-tension that they have a more sophisticated understanding of physical interactions involved in the task. Though only woodpecker finches solved the initial task, all failed in the follow-ing transfer task. Furthermore, the proportion of successful woodpecker finches in the initial task was not significantly higher than for small tree finches.
The fact that small tree finches performed equally well or better than woodpecker finches in other tasks testing physical cognition (Chapter 1) provides further reason to be cautious in interpreting the poor performance of the small tree finches in the trap tube tasks.
The results parallel the findings of the previous woodpecker finch study (Tebbich and Bshary 2004) in the sense that woodpecker finches learned to avoid a trap but did not seem able to appreciate the function of a trap even after being given numerous opportunities to do so. We also did not observe that tool-experienced woodpecker finches performed better in the trap tube with the pre-inserted stick (Experiment 2). Indeed, the only woodpecker finches that were able to solve the first stage of this task were non-tool-users.
One tool-using woodpecker finch solved the initial task when allowed to insert a stick tool (Experiment 3) but none were able to do so using the pre-inserted tool (Experiments 1 and 2). This might be due to the more natural context of task presentation in Experiment 3—even chimpanzees and orangutans perform better in the original trap tube task when allowed to apply a “species-specific tool-using action” (Mulcahy and Call 2006, p. 194). However, the failure of this subject in the transfer task indicates that it also probably solved the initial task using an arbitrary cue that was altered in the transfer condition.
When one considers the relative uniformity in the natural tool-use of woodpecker finches whereby tools are used only in one context and in one way, it makes sense deploying tools in this species does not necessitate generalization of physical interactions—simple situation-specific rules probably suffice to get the job done. However, the low number of individuals that were able to solve the initial task even using simple context-specific rules, shows that an easier task would be needed to detect a significant species difference with such a low sample size. One possibility would be to present them with a task which allows direct movement of the food with the beak as in liedtke et al. (2010).
The fact that large-brained rooks are able to extract generalized rules in a task where small tree finches do not even succeed in successfully applying a simple procedural rule in first stage of the task suggests a qualitative difference in the cognition of these non-tool-using species. Likewise, the comparison of the performance of tool-using woodpecker finches and New Caledonian crows in the two-trap tube task that was operated with a freely manipulable tool also shows that larger brained New Caledonian crows have a higher propensity to form a general rule while woodpecker finches only can form a situation-specific rule. Finally, one chimpanzee that was tested in a further variation of the two-trap tube paradigm was able to solve a series of transfer tasks that had no simple perceptual cue in common. The summary of these results suggests that large brain size might be a better predictor of the ability to form
a generalized rule pertaining to the physical properties of the task than tool-use. Only the performance of parrots does not fit in with this interpretation. Parrots are relatively large-brained but they failed to even extract a simple procedural rule to solve the initial single trap tube task with a pre-inserted tool when given between 100 and 200 trials to do so. Even woodpecker finches with their presumably smaller proportional brain sizes solved the initial task using a simple rule (see liedtke et al. 2010 for a detailed discussion).
Woodpecker finch tool-use is characterized by selectivity, modification, and high frequency in natural populations and furthermore “is not a stereotypic behavioural pattern, but is open to modification by learning” (Tebbich and Bshary 2004, p. 696). Nevertheless, all studies to date which have investigated cognition related to tool-use in woodpecker finches have failed to provide any evidence that this species possesses sophisticated physical cognitive abilities, that they use mental representation and planning in problem-solving related to tool-use (Teb-bich and Bshary 2004; Chapter 1) nor have they yielded evidence that woodpecker finches must learn this seemingly complex technique from other conspecifics (Tebbich et al. 2001).
In woodpecker finches, simple cognitive solutions appear to suffice for the ontogenetic de-velopment of tool-use and for its deployment. In particular, trial-and-error learning appears vital to the species in acquiring tool-using skills in ontogeny (Tebbich et al. 2001) and also in solving a battery of physical problems.
ACKNOWLEDGEMENTS
IT was supported by the German research foundation (DFG, Project Nr. TE628/1-1) and ST by the Austrian Science Fund (FWF, Project Nr. V95-B17). The experiments comply with the current laws of the country in which they were performed. We are thankful to the Charles Darwin research Station for support and TAME for reduced ticket fares. We are also grate-ful to Dr. Birgit Fessl for support in all facets of this study. Caroline raby, Viviana Morales, Mari Cruz Jaramillo, Tania Quisingo Chiza, Paola Buitron lopez, and Eduardo Sandoval provided valuable field assistance and help with experiments. Patrick Meidl provided vital support in organizing vast amounts of data. Thanks to Andy Burnley for constructing ex-perimental apparatus and to Sue-Anne Zollinger for helping to make the figure depicting the experimental apparatus.
The ability to unlearn a previously established association is an important component of behavioural flexibility and may vary according to species ecology. Previously, two closely related, sympatric Darwin’s finches, the woodpecker finch (Cactospiza pallida) and the small tree finch (Camarhynchus parvulus), were found to differ significantly in their ability to solve a novel operant task and in reversal learning, a test of behavioural flexibility. Small tree finches outperformed woodpecker finches in reversal learning but performed worse in the operant task. We attributed this difference to the habit of woodpecker finches to engage in long bouts of energetic pecking at a substrate during extractive foraging. Persistently repeating one action without reward could favour performance in operant tasks but also limit behavioural flexibility. Here, we tested whether perseverance is the reason for woodpecker finches’ depressed reversal learning performance. Two modified reversal conditions allowed the disentanglement of two sources of error in reversal learning: perseverant choice of the previously rewarded stimulus and failure to respond to the previously non-rewarded stimulus. Our prediction was that woodpecker finches should make more errors of perseverance than small tree finches.
Since performance differences could also be influenced by reaction to novelty, we compared neophilic and neophobic responses in woodpecker finches and small tree finches and related them to reversal learning proficiency. We found no significant species difference in reversal learning under these new conditions, suggesting that the observed difference from the previous study is not attributable to high perseverance in woodpecker finches. However, woodpecker finches were less neophobic and there was some indication for a negative correlation between neophobia and reversal learning performance.
Abstract
l
earning enables animals to fine-tune their behavioural response to their respective environment. Thus, learning abilities may vary with feeding ecology, even between closely related species (e.g. Garcia et al. 1974; Laverty and Plowright 1988; Ratcliffe et al.2003) and between populations from different habitats (roth et al. 2010). The complexity and the variability of an environment is thought to be one of the key factors that determine the optimal rate of learning (Godfrey-Smith 2001). Species or populations that live in complex or fluctuating environments must react more readily to changes and should therefore be more
INVESTIGATING THE RELATIONSHIP BETWEEN FORAGING AND LEARNING ABILITIES IN TWO SPECIES OF DARWIN’S FINCHES
Tebbich, S., Stankewitz S. and Teschke, I.
(manuscript)
flexible than individuals from stable environments (Day et al. 2003; Jones 2005; Robinson 1990; Shettleworth 1998; Tomasello and Call 1997). Even closely related species that live in the same habitat might experience the predictability of relevant resources differently depending on their foraging ecology. The ability to inhibit previously successful responses is one factor that could enhance flexibility under changing conditions and the most widespread paradigm utilized to test this aspect of flexibility is that of reversal learning. In reversal learning tests, subjects first must learn a simple discrimination and once a set learning criterion is reached, the reward contingencies are reversed in a second learning phase. A wide range of species have been tested in the serial reversal paradigm in an attempt to find a comparable measure of learning performance (Bittermann 1965; Davey 1989; Mackintosh and Holgate 1969; Mackintosh et al. 1968). More recently this experimental paradigm has been used to compare closely related species and different populations of one species to test more precise predictions about the relationship of behavioural flexibility and the complexity of the environment (Bond et al. 2007; Day et al. 1999; roth et al. 2010).
However, differences in learning abilities can also be due to differences in reactions to novelty, which can either directly or indirectly influence learning. For instance a novel stim-ulus could evoke aversion or attraction during an interaction and thus influence learning per-formance directly. Seferta et al. (2001) discovered that feral pigeons (Columba livia), which are afraid of novel stimuli, are slower at learning an operant task than the less fearful Zenaida doves (Zenaida aurita). Furthermore, Webster and lefebvre (2001) found the same pattern when comparing the relationship between problem-solving and fearful reactions towards novel stimuli in five different species of passeriformes and columbiformes and also between individuals. However, not only the aversion of novelty but also the attraction to novelty can influence learning. For instance Auersperg et al. (submitted) found that the exceptionally explorative Keas (Nestor notabilis) were faster at finding solutions to a problem-solving task with four possible solutions than the less explorative New Caledonian crows (Corvus moneduloides).
Aversion to novelty is thought to be driven by neophobia which is defined as avoidance of a food item, object or place only because it is novel, whereas the attraction to novelty is thought to be driven by neophilia, defined as the spontaneous attraction of an animal to a food item, object or place simply because it is novel (Thorpe 1956). Some theoretical models consider neophobia and neophilia as opposing ends of one behavioural continuum (Hogan 1965; Thorpe 1956) while others see them as two distinct, but related motivational systems which can be aroused simultaneously (Chance and Meade 1955; Greenberg and Mettke-Hofmann 2001; Hughes 1997; Montgomery 1955; russell 1973). Most importantly it is believed that the two motivational systems are driven by different selective forces and may therefore vary independently of each other, depending on species ecology. While it is surmised that neophobia is driven by the costs of exploration, such as predation, neophilia is thought to be driven by the benefits of exploration such as the discovery of novel resources (Greenberg and Mettke-Hofmann 2001).
In the current study, our goal was to test predictions about the relationship between ex-tractive foraging, and reversal learning and reaction to novelty in two species of Darwin’s Finches, the woodpecker finch (Cactospiza pallida) and the small tree finch (Camarhynchus parvulus). A previous study showed that these two closely related species differ significantly in reversal learning and in learning an operant task (Chapter 1): small tree finches outper-formed woodpecker finches in a classical colour discrimination reversal paradigm (Chapter 1, Experiment 1) whereas woodpecker finches outperformed small tree finches in a novel operant task (Chapter 1, Experiment 4). We attributed these findings to the different feeding ecologies of the species, especially to the persistent foraging style of woodpecker finches.
The sympatric woodpecker finch and the small tree finch have a similar diet composition and also utilize similar feeding techniques (Tebbich et al. 2004) but show a pronounced differ-ence in their proportion of extractive foraging. In contrast to small tree finches, woodpecker finches often engage in long bouts of energetic pecking at wood and probing into crevices without immediately obtaining their prey. This foraging strategy requires high levels of per-severance, and this, in turn, requires that a subject continues with an activity, even when it is not rewarded.
This dissimilarity in feeding ecology could be crucial in understanding the disparity in the performance of woodpecker finches and small tree finches in the learning experiments.
On one hand, high perseverance can lead to success in an operant task, because frequent confrontation with an object increases the probability of accidentally making the move-ments required to solve the task. On the other hand, fast reversal learning requires just the opposite: an animal must be sensitive to not being rewarded for a previously profitable be-haviour and respond by changing its bebe-haviour. Extractive foraging that requires persever-ance may therefore come at a cost, namely it might impede learning flexibility because high perseverance is likely to be associated with a low sensitivity to absence of reinforcement (Tebbich et al. 2010; Chapter 1). If this is the case, we could expect a disparity in flexibility between extractive and non-extractive foragers that is widespread among different taxa. An effect of extractive foraging on cognitive abilities has been demonstrated in a comparison between predominantly gum feeding marmosets (Callithrix jacchus) and more insectivorous tamarins (Saguinus oedipus): in a self-control paradigm the marmosets waited significantly longer for food than tamarins, which indicates that the patience needed to feed on gums may have selected for a more general ability to cope with delayed gratification (Stevens et al.
2005). Extractive foraging does not only seem to influence cognitive abilities but also reac-tion to novelty. In a comparative study of parrots Mettke-Hofmann et al. (2002) found that the duration of exploration correlated positively with the proportion of nuts in the diet. In contrast, extractive foraging correlated negatively the propensity to approach novel objects in Darwin’s finches (Tebbich et al. 2009).
In the current study we aimed to test whether the difference in reversal learning found between the two species in Chapter 1 is indeed a consequence of a difference in sensitivity to a change in reward contingency (sensitivity to the absence of reinforcement) utilizing two
modifications of the classical reversal learning paradigm.
Following Tait and Brown (2007), we applied a reversal design that allowed the disen-tanglement of two possible sources of error in classical reversal learning which cannot be distinguished in the classical paradigm: in the “Perseverance” reversal (hereafter P reversal condition), the formerly rewarded colour ceases to be rewarded and a newly introduced col-our becomes the rewarded stimulus while in the “learned Non-reward” reversal (hereafter lNr reversal condition), the formerly unrewarded colour is now rewarded and a newly introduced colour becomes the unrewarded stimulus (Figure 3.1).
In light of their extractive foraging style, we expected woodpecker finches to be less sen-sitive to a change in which a formerly rewarded stimulus is not rewarded anymore. If this is the reason for the poor performance of woodpecker finches, they should make more errors in the P reversal than in the lNr reversal and also make more errors in the P reversal than small tree finches.
As learning can also be influenced by reaction to novelty, another objective was to inves-tigate whether small tree finches and woodpecker finches differ in their novelty responses, whether individual reaction to novelty was consistent across trials and how learning and reaction to novelty are related. A consistent inter-individual reaction to novelty indicates that this trait may be related to an individual’s coping styles. If this is so, differences in reaction to novelty may not only directly influence reversal learning via attraction or aversion to nov-el stimuli but also indirectly via the coping style since studies on animal personality indicate that an animal’s reaction to novel stimuli is likely to be correlated to the learning abilities of the species. For instance, Drent and Marchetti (1999) and Verbeek et al. (1994) found that in great tits (Parus major), slow explorers changed a previously established foraging habit more quickly than fast explorers, indicating that responses to novelty and reversal learning (i. e. learning of changed reward contingencies in a discrimination task) are co-varying traits within this species.
The mechanism underlying the relationship between reversal learning and reaction to novelty is not known. Verbeek et al. (1994) and Drent and Marchetti (1999) argue that fast explorers were quicker to establish a foraging habit, but slower at changing an established habit, whereas slow explorers quickly extended their search to other places when they did not find food readily at established sites. These are, however, two different (but not mutually exclusive) explanations that are related to the previously described components of reversal learning namely perseverance in choosing the previously rewarded stimulus and difficulty in learning that a previously non-rewarded stimulus is now rewarded. The data from the lNr and P reversal conditions in combination with the data on novelty reactions allowed us to test whether more neophobic birds are more flexible because they abandon a learned feeding habit more quickly (P reversal, learn that the previously rewarded stimulus is not rewarded anymore) or because they learn to choose a previously unrewarded stimulus more quickly (i.e. “remain alert to stimuli in the known environment” Verbeek et al. 1994, p. 119). To as-sess potential species differences in novelty reactions we compared neophilic and neophobic
responses towards novel objects in woodpecker finches and small tree finches and examined the relationship of these results with those of the learning experiments.
METHODS
Study area, subjects and housing
Study area This study was conducted on Santa Cruz, which is one of the larger, central islands (986 km2) of the Galápagos archipelago. Birds for the experiments were caught in the so-called Scalesia zone, located at an elevation of between 300m and 650m. This veg-etation zone is characterized by the eponymous tree Scalesia pedunculata of the Asteraceae family which forms an evergreen, lush, moist forest. The branches and trunks of these trees are densely covered with epiphytic moss, lichens, ferns and bromeliads. Prey arthropods are abundant throughout the year in this area and can be found primarily in moss, on leaves and under bark (Tebbich et al. 2002).
Study species The study species were two closely related species of Darwin’s Finches:
the woodpecker finch and the small tree finch. They occur sympatrically in the Scalesia zone on Santa Cruz Island. Both species are members of the tree finch clade and within this group are members of a guild of mainly insectivorous tree foragers.
Woodpecker finches (~20g) have an elongated but powerful beak that is suitable for prob-ing moss patches and peckprob-ing into wood and under bark to gain access to their prey. This
Woodpecker finches (~20g) have an elongated but powerful beak that is suitable for prob-ing moss patches and peckprob-ing into wood and under bark to gain access to their prey. This